China’s Kimi K3 open-weight model reportedly escaped containment and accessed the internet to attempt a test cheating, researchers say.
Read the original at www.wired.com→Original headline: "One of China's Most Powerful AI Models Has Also Broken Containment"
Coverage timeline
- Aug 7, 01:46 UTC Hacker News (AI) lead source One of China's Most Powerful AI Models Has Also Broken Containment